Debt covenant slack and ex-post conditional accounting conservatism

Abstract

This study examines how conditional accounting conservatism varies after debt contracts are executed (ex-post conditional conservatism) depending on the closeness to violations of debt covenants, the costs of covenant violations, and bank monitoring. I hypothesise that borrowers with lower debt-covenant slack have an incentive to avoid covenant violations and will have lower ex-post conditional conservatism levels than those with higher slack. I find that borrowers with both lower and higher slack increase their ex-post conditional conservatism, but that the magnitude of this increase in conservatism is positively associated with the level of covenant slack. Further, this positive relationship between the ex-post change in conditional conservatism and the debt-covenant slack is more pronounced when the costs of the debt-covenant breach are greater, and is less pronounced when lenders have stronger incentives to monitor borrowers. This study provides evidence that the ex-post change in conditional conservatism is affected by the costs of a covenant breach and bank monitoring.

Keywords:

• debt covenant hypothesis
• conditional accounting conservatism
• asymmetric timeliness
• accruals

Notes

1 Roberts and Sufi (2009) demonstrate that 90% of loans longer than one year are renegotiated before less than half of the original stated maturity has elapsed.

2 As Nikolaev (2010) argues, one can also consider these residuals as instrumental variables; like instruments, they correlate with the covenant slack, although by construction they are uncorrelated with other variables that potentially confound the relationship.

3 A negative relationship between ex-ante conditional conservatism and covenant slack is possible if lenders believe conservative borrowers have greater flexibility to reduce conditional conservatism after borrowing to avoid a covenant breach and establish tight slack for borrowers with higher ex-ante conditional conservatism. A positive relationship between ex-ante conditional conservatism and covenant slack is possible if lenders believe conservative borrowers are less risky and establish high slack for borrowers with higher ex-ante conditional conservatism. My results indicate that no relationship exists between covenant slack and ex-ante conditional conservatism, suggesting both effects nullify each other.

4 I also divide the slack by the net-worth covenant threshold as a robustness check, and my results remain consistent.

5 In this study, I examine whether conditional conservatism changes after borrowing, and conditional conservatism is measured using financial statements. Hence, conservatism at the moment of obtaining the loan (or debt contracts) is measured using the financial statements available at the time of the contract. I assume that annual financial statements are available three months after the fiscal year-end; therefore, I measure conservatism at the moment of the debt contract using financial statements available at the deal date, and call this year t, or ex-ante conditional conservatism. I measure conditional conservatism one year after each fiscal year-end and call this year t + 1, or ex-post conservatism. Slack is also measured using financials available at the time of the contract, but more precisely before the contract, as the closeness to a covenant violation at year t will affect ex-post (year t + 1) conditional conservatism.

6 I also use the modified Jones (1991) discretionary accruals (Dechow et al. 1995) and obtain qualitatively the same results.

7 I obtained covenant violation data from Amir Sufi’s website following Nini et al. (2012).

8 The difference in coefficients is significant at the 10% level.

9 For example, lenders may believe a future increase in conservatism will be limited under high ex-ante conservatism.

10 The results using raw slack are qualitatively the same.

11 I re-estimate the regression using CFO from the cash flow statement and obtain similar results. The coefficients of interest are 0.354 in Column (1) and 0.291 in Column (3), with a 5% or higher statistical significance.